Text Generation
Transformers
Safetensors
llama
llama2
svd
compression
safety
interpretability
conversational
text-generation-inference
Instructions to use Jeesup/svd-safety-l2_swift_remove50 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Jeesup/svd-safety-l2_swift_remove50 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Jeesup/svd-safety-l2_swift_remove50") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Jeesup/svd-safety-l2_swift_remove50") model = AutoModelForCausalLM.from_pretrained("Jeesup/svd-safety-l2_swift_remove50", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Jeesup/svd-safety-l2_swift_remove50 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Jeesup/svd-safety-l2_swift_remove50" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-l2_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Jeesup/svd-safety-l2_swift_remove50
- SGLang
How to use Jeesup/svd-safety-l2_swift_remove50 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Jeesup/svd-safety-l2_swift_remove50" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-l2_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Jeesup/svd-safety-l2_swift_remove50" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-l2_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Jeesup/svd-safety-l2_swift_remove50 with Docker Model Runner:
docker model run hf.co/Jeesup/svd-safety-l2_swift_remove50
Download compression.json from Jeesup/svd-safety-l2_swift_remove50: direct link, hf CLI and curl.
- Browser
- Download file 15.2 kB
-
https://huggingface.co/Jeesup/svd-safety-l2_swift_remove50/resolve/main/compression.json
- Command line
-
hf download hf://Jeesup/svd-safety-l2_swift_remove50/compression.json
-
curl -L -o compression.json https://huggingface.co/Jeesup/svd-safety-l2_swift_remove50/resolve/main/compression.json
15.2 kB
| { | |
| "source": "/workspace/svd_safety/assets/models/Llama-2-7b-chat-hf", | |
| "upstream": "hiahei/Swift-SVD (compression) + AIoT-MLSys-Lab/SVD-LLM utils/LoRA.py (recovery)", | |
| "stage": "finetuned", | |
| "ratio_removed": 0.5, | |
| "internal_ratio_kept": 0.5, | |
| "from_pt": "/workspace/svd_safety/results/swift_ft/cells/l2_swift_remove50/finetune_v/merge.pt", | |
| "mean_rel_err": 0.5770199772502694, | |
| "max_rel_err": 0.6914882659912109, | |
| "param_fraction": 0.4998633824481865, | |
| "ranks": { | |
| "0.q_proj": 1212, | |
| "0.k_proj": 1202, | |
| "0.v_proj": 1147, | |
| "0.o_proj": 1160, | |
| "0.gate_proj": 1526, | |
| "0.up_proj": 1527, | |
| "0.down_proj": 1707, | |
| "1.q_proj": 1121, | |
| "1.k_proj": 1118, | |
| "1.v_proj": 1078, | |
| "1.o_proj": 1075, | |
| "1.gate_proj": 1509, | |
| "1.up_proj": 1509, | |
| "1.down_proj": 1775, | |
| "2.q_proj": 1022, | |
| "2.k_proj": 1022, | |
| "2.v_proj": 1006, | |
| "2.o_proj": 983, | |
| "2.gate_proj": 1487, | |
| "2.up_proj": 1487, | |
| "2.down_proj": 1404, | |
| "3.q_proj": 1029, | |
| "3.k_proj": 1029, | |
| "3.v_proj": 1022, | |
| "3.o_proj": 998, | |
| "3.gate_proj": 1490, | |
| "3.up_proj": 1490, | |
| "3.down_proj": 1433, | |
| "4.q_proj": 1052, | |
| "4.k_proj": 1053, | |
| "4.v_proj": 1044, | |
| "4.o_proj": 1026, | |
| "4.gate_proj": 1496, | |
| "4.up_proj": 1496, | |
| "4.down_proj": 1492, | |
| "5.q_proj": 1031, | |
| "5.k_proj": 1032, | |
| "5.v_proj": 1024, | |
| "5.o_proj": 1011, | |
| "5.gate_proj": 1492, | |
| "5.up_proj": 1492, | |
| "5.down_proj": 1461, | |
| "6.q_proj": 1047, | |
| "6.k_proj": 1048, | |
| "6.v_proj": 1045, | |
| "6.o_proj": 1033, | |
| "6.gate_proj": 1497, | |
| "6.up_proj": 1496, | |
| "6.down_proj": 1500, | |
| "7.q_proj": 1025, | |
| "7.k_proj": 1025, | |
| "7.v_proj": 1023, | |
| "7.o_proj": 1013, | |
| "7.gate_proj": 1492, | |
| "7.up_proj": 1491, | |
| "7.down_proj": 1465, | |
| "8.q_proj": 1033, | |
| "8.k_proj": 1034, | |
| "8.v_proj": 1033, | |
| "8.o_proj": 1026, | |
| "8.gate_proj": 1494, | |
| "8.up_proj": 1494, | |
| "8.down_proj": 1486, | |
| "9.q_proj": 1017, | |
| "9.k_proj": 1018, | |
| "9.v_proj": 1017, | |
| "9.o_proj": 1014, | |
| "9.gate_proj": 1491, | |
| "9.up_proj": 1490, | |
| "9.down_proj": 1459, | |
| "10.q_proj": 1009, | |
| "10.k_proj": 1010, | |
| "10.v_proj": 1009, | |
| "10.o_proj": 1008, | |
| "10.gate_proj": 1489, | |
| "10.up_proj": 1489, | |
| "10.down_proj": 1445, | |
| "11.q_proj": 1008, | |
| "11.k_proj": 1009, | |
| "11.v_proj": 1011, | |
| "11.o_proj": 1007, | |
| "11.gate_proj": 1489, | |
| "11.up_proj": 1488, | |
| "11.down_proj": 1445, | |
| "12.q_proj": 1004, | |
| "12.k_proj": 1004, | |
| "12.v_proj": 1006, | |
| "12.o_proj": 1004, | |
| "12.gate_proj": 1487, | |
| "12.up_proj": 1487, | |
| "12.down_proj": 1441, | |
| "13.q_proj": 1007, | |
| "13.k_proj": 1008, | |
| "13.v_proj": 1010, | |
| "13.o_proj": 1009, | |
| "13.gate_proj": 1489, | |
| "13.up_proj": 1489, | |
| "13.down_proj": 1450, | |
| "14.q_proj": 1004, | |
| "14.k_proj": 1005, | |
| "14.v_proj": 1006, | |
| "14.o_proj": 1007, | |
| "14.gate_proj": 1487, | |
| "14.up_proj": 1487, | |
| "14.down_proj": 1446, | |
| "15.q_proj": 1005, | |
| "15.k_proj": 1006, | |
| "15.v_proj": 1007, | |
| "15.o_proj": 1010, | |
| "15.gate_proj": 1488, | |
| "15.up_proj": 1487, | |
| "15.down_proj": 1452, | |
| "16.q_proj": 1005, | |
| "16.k_proj": 1006, | |
| "16.v_proj": 1008, | |
| "16.o_proj": 1014, | |
| "16.gate_proj": 1489, | |
| "16.up_proj": 1489, | |
| "16.down_proj": 1459, | |
| "17.q_proj": 1002, | |
| "17.k_proj": 1003, | |
| "17.v_proj": 1006, | |
| "17.o_proj": 1007, | |
| "17.gate_proj": 1488, | |
| "17.up_proj": 1487, | |
| "17.down_proj": 1456, | |
| "18.q_proj": 1000, | |
| "18.k_proj": 1000, | |
| "18.v_proj": 1005, | |
| "18.o_proj": 1005, | |
| "18.gate_proj": 1487, | |
| "18.up_proj": 1487, | |
| "18.down_proj": 1456, | |
| "19.q_proj": 994, | |
| "19.k_proj": 995, | |
| "19.v_proj": 1000, | |
| "19.o_proj": 1000, | |
| "19.gate_proj": 1486, | |
| "19.up_proj": 1486, | |
| "19.down_proj": 1447, | |
| "20.q_proj": 994, | |
| "20.k_proj": 994, | |
| "20.v_proj": 999, | |
| "20.o_proj": 1003, | |
| "20.gate_proj": 1486, | |
| "20.up_proj": 1486, | |
| "20.down_proj": 1450, | |
| "21.q_proj": 991, | |
| "21.k_proj": 991, | |
| "21.v_proj": 998, | |
| "21.o_proj": 997, | |
| "21.gate_proj": 1485, | |
| "21.up_proj": 1485, | |
| "21.down_proj": 1446, | |
| "22.q_proj": 991, | |
| "22.k_proj": 991, | |
| "22.v_proj": 998, | |
| "22.o_proj": 1002, | |
| "22.gate_proj": 1485, | |
| "22.up_proj": 1485, | |
| "22.down_proj": 1447, | |
| "23.q_proj": 989, | |
| "23.k_proj": 989, | |
| "23.v_proj": 999, | |
| "23.o_proj": 999, | |
| "23.gate_proj": 1485, | |
| "23.up_proj": 1485, | |
| "23.down_proj": 1446, | |
| "24.q_proj": 988, | |
| "24.k_proj": 989, | |
| "24.v_proj": 997, | |
| "24.o_proj": 1001, | |
| "24.gate_proj": 1485, | |
| "24.up_proj": 1485, | |
| "24.down_proj": 1445, | |
| "25.q_proj": 989, | |
| "25.k_proj": 988, | |
| "25.v_proj": 1000, | |
| "25.o_proj": 1006, | |
| "25.gate_proj": 1485, | |
| "25.up_proj": 1485, | |
| "25.down_proj": 1449, | |
| "26.q_proj": 989, | |
| "26.k_proj": 989, | |
| "26.v_proj": 1000, | |
| "26.o_proj": 1007, | |
| "26.gate_proj": 1485, | |
| "26.up_proj": 1486, | |
| "26.down_proj": 1451, | |
| "27.q_proj": 990, | |
| "27.k_proj": 990, | |
| "27.v_proj": 1001, | |
| "27.o_proj": 1006, | |
| "27.gate_proj": 1486, | |
| "27.up_proj": 1486, | |
| "27.down_proj": 1457, | |
| "28.q_proj": 991, | |
| "28.k_proj": 991, | |
| "28.v_proj": 1003, | |
| "28.o_proj": 1013, | |
| "28.gate_proj": 1486, | |
| "28.up_proj": 1487, | |
| "28.down_proj": 1462, | |
| "29.q_proj": 992, | |
| "29.k_proj": 992, | |
| "29.v_proj": 1003, | |
| "29.o_proj": 1017, | |
| "29.gate_proj": 1487, | |
| "29.up_proj": 1487, | |
| "29.down_proj": 1469, | |
| "30.q_proj": 1003, | |
| "30.k_proj": 1002, | |
| "30.v_proj": 1016, | |
| "30.o_proj": 1026, | |
| "30.gate_proj": 1491, | |
| "30.up_proj": 1492, | |
| "30.down_proj": 1525, | |
| "31.q_proj": 1234, | |
| "31.k_proj": 1235, | |
| "31.v_proj": 1247, | |
| "31.o_proj": 1281, | |
| "31.gate_proj": 1545, | |
| "31.up_proj": 1547, | |
| "31.down_proj": 2018 | |
| }, | |
| "rel_err": { | |
| "0.q_proj": 0.1519796997308731, | |
| "0.k_proj": 0.1192033439874649, | |
| "0.v_proj": 0.38923344016075134, | |
| "0.o_proj": 0.3858165740966797, | |
| "0.gate_proj": 0.500773549079895, | |
| "0.up_proj": 0.5168083906173706, | |
| "0.down_proj": 0.5532068014144897, | |
| "1.q_proj": 0.16660481691360474, | |
| "1.k_proj": 0.1496836543083191, | |
| "1.v_proj": 0.5328048467636108, | |
| "1.o_proj": 0.4690547585487366, | |
| "1.gate_proj": 0.48617124557495117, | |
| "1.up_proj": 0.5469284653663635, | |
| "1.down_proj": 0.5976937413215637, | |
| "2.q_proj": 0.35211285948753357, | |
| "2.k_proj": 0.3386683166027069, | |
| "2.v_proj": 0.5837520956993103, | |
| "2.o_proj": 0.6466013193130493, | |
| "2.gate_proj": 0.556005597114563, | |
| "2.up_proj": 0.5962874293327332, | |
| "2.down_proj": 0.6686410903930664, | |
| "3.q_proj": 0.45740288496017456, | |
| "3.k_proj": 0.45834606885910034, | |
| "3.v_proj": 0.5986074209213257, | |
| "3.o_proj": 0.6326659917831421, | |
| "3.gate_proj": 0.613806426525116, | |
| "3.up_proj": 0.6380705833435059, | |
| "3.down_proj": 0.6272472739219666, | |
| "4.q_proj": 0.4375986158847809, | |
| "4.k_proj": 0.4429137408733368, | |
| "4.v_proj": 0.5889207720756531, | |
| "4.o_proj": 0.6280702352523804, | |
| "4.gate_proj": 0.6121411919593811, | |
| "4.up_proj": 0.6420316100120544, | |
| "4.down_proj": 0.6181107759475708, | |
| "5.q_proj": 0.4545028805732727, | |
| "5.k_proj": 0.45424583554267883, | |
| "5.v_proj": 0.600082278251648, | |
| "5.o_proj": 0.636687159538269, | |
| "5.gate_proj": 0.6082773804664612, | |
| "5.up_proj": 0.6440009474754333, | |
| "5.down_proj": 0.6173216700553894, | |
| "6.q_proj": 0.4331511855125427, | |
| "6.k_proj": 0.44137337803840637, | |
| "6.v_proj": 0.5765433311462402, | |
| "6.o_proj": 0.599454939365387, | |
| "6.gate_proj": 0.6015103459358215, | |
| "6.up_proj": 0.6436546444892883, | |
| "6.down_proj": 0.6076619625091553, | |
| "7.q_proj": 0.4461621642112732, | |
| "7.k_proj": 0.4536467492580414, | |
| "7.v_proj": 0.5764104127883911, | |
| "7.o_proj": 0.6230320930480957, | |
| "7.gate_proj": 0.599027693271637, | |
| "7.up_proj": 0.6438522934913635, | |
| "7.down_proj": 0.6086878180503845, | |
| "8.q_proj": 0.45493605732917786, | |
| "8.k_proj": 0.4706614911556244, | |
| "8.v_proj": 0.5578523874282837, | |
| "8.o_proj": 0.6163948178291321, | |
| "8.gate_proj": 0.5954751968383789, | |
| "8.up_proj": 0.6374110579490662, | |
| "8.down_proj": 0.5979208946228027, | |
| "9.q_proj": 0.4763244092464447, | |
| "9.k_proj": 0.48309847712516785, | |
| "9.v_proj": 0.5613107085227966, | |
| "9.o_proj": 0.6197980642318726, | |
| "9.gate_proj": 0.5919429659843445, | |
| "9.up_proj": 0.632355809211731, | |
| "9.down_proj": 0.6026204228401184, | |
| "10.q_proj": 0.44654256105422974, | |
| "10.k_proj": 0.459806352853775, | |
| "10.v_proj": 0.5551357865333557, | |
| "10.o_proj": 0.5909467339515686, | |
| "10.gate_proj": 0.5873189568519592, | |
| "10.up_proj": 0.6278164982795715, | |
| "10.down_proj": 0.6019299626350403, | |
| "11.q_proj": 0.4331582188606262, | |
| "11.k_proj": 0.44750073552131653, | |
| "11.v_proj": 0.539775550365448, | |
| "11.o_proj": 0.5791270136833191, | |
| "11.gate_proj": 0.5880950689315796, | |
| "11.up_proj": 0.62898188829422, | |
| "11.down_proj": 0.6085322499275208, | |
| "12.q_proj": 0.47685113549232483, | |
| "12.k_proj": 0.47473812103271484, | |
| "12.v_proj": 0.5630918741226196, | |
| "12.o_proj": 0.62078458070755, | |
| "12.gate_proj": 0.5933976769447327, | |
| "12.up_proj": 0.6293315887451172, | |
| "12.down_proj": 0.6097376346588135, | |
| "13.q_proj": 0.4816206395626068, | |
| "13.k_proj": 0.5003007650375366, | |
| "13.v_proj": 0.5929780602455139, | |
| "13.o_proj": 0.60856032371521, | |
| "13.gate_proj": 0.5962281823158264, | |
| "13.up_proj": 0.6323710680007935, | |
| "13.down_proj": 0.6194044947624207, | |
| "14.q_proj": 0.45993104577064514, | |
| "14.k_proj": 0.4645041525363922, | |
| "14.v_proj": 0.5858851671218872, | |
| "14.o_proj": 0.6054980754852295, | |
| "14.gate_proj": 0.5965513586997986, | |
| "14.up_proj": 0.6316479444503784, | |
| "14.down_proj": 0.6229026913642883, | |
| "15.q_proj": 0.4835875630378723, | |
| "15.k_proj": 0.48482608795166016, | |
| "15.v_proj": 0.59104323387146, | |
| "15.o_proj": 0.6179000735282898, | |
| "15.gate_proj": 0.6003255248069763, | |
| "15.up_proj": 0.6353552341461182, | |
| "15.down_proj": 0.63531893491745, | |
| "16.q_proj": 0.4797825813293457, | |
| "16.k_proj": 0.48336681723594666, | |
| "16.v_proj": 0.620274007320404, | |
| "16.o_proj": 0.6427569389343262, | |
| "16.gate_proj": 0.5996921062469482, | |
| "16.up_proj": 0.632495105266571, | |
| "16.down_proj": 0.6475937366485596, | |
| "17.q_proj": 0.5249422788619995, | |
| "17.k_proj": 0.5217642784118652, | |
| "17.v_proj": 0.635169506072998, | |
| "17.o_proj": 0.6561335921287537, | |
| "17.gate_proj": 0.6028484106063843, | |
| "17.up_proj": 0.6339511871337891, | |
| "17.down_proj": 0.6465868353843689, | |
| "18.q_proj": 0.5513262748718262, | |
| "18.k_proj": 0.5563322901725769, | |
| "18.v_proj": 0.6416341066360474, | |
| "18.o_proj": 0.6792891025543213, | |
| "18.gate_proj": 0.6122872233390808, | |
| "18.up_proj": 0.6402170658111572, | |
| "18.down_proj": 0.6462416052818298, | |
| "19.q_proj": 0.5357663035392761, | |
| "19.k_proj": 0.530946671962738, | |
| "19.v_proj": 0.6263485550880432, | |
| "19.o_proj": 0.6681627631187439, | |
| "19.gate_proj": 0.6159399151802063, | |
| "19.up_proj": 0.6408361196517944, | |
| "19.down_proj": 0.6481027603149414, | |
| "20.q_proj": 0.5267692804336548, | |
| "20.k_proj": 0.5284019708633423, | |
| "20.v_proj": 0.6414449214935303, | |
| "20.o_proj": 0.6711487174034119, | |
| "20.gate_proj": 0.6140052080154419, | |
| "20.up_proj": 0.6376996636390686, | |
| "20.down_proj": 0.6590219140052795, | |
| "21.q_proj": 0.5438850522041321, | |
| "21.k_proj": 0.5463284850120544, | |
| "21.v_proj": 0.6339064836502075, | |
| "21.o_proj": 0.6861968040466309, | |
| "21.gate_proj": 0.61533522605896, | |
| "21.up_proj": 0.6377876400947571, | |
| "21.down_proj": 0.65590900182724, | |
| "22.q_proj": 0.5565619468688965, | |
| "22.k_proj": 0.5612902045249939, | |
| "22.v_proj": 0.6452638506889343, | |
| "22.o_proj": 0.6826830506324768, | |
| "22.gate_proj": 0.6154788732528687, | |
| "22.up_proj": 0.6370535492897034, | |
| "22.down_proj": 0.6531917452812195, | |
| "23.q_proj": 0.5894484519958496, | |
| "23.k_proj": 0.5889238715171814, | |
| "23.v_proj": 0.6437049508094788, | |
| "23.o_proj": 0.6743727326393127, | |
| "23.gate_proj": 0.6192212700843811, | |
| "23.up_proj": 0.6390761137008667, | |
| "23.down_proj": 0.6452059745788574, | |
| "24.q_proj": 0.5363700985908508, | |
| "24.k_proj": 0.5404852628707886, | |
| "24.v_proj": 0.6358501315116882, | |
| "24.o_proj": 0.6789605617523193, | |
| "24.gate_proj": 0.6206092238426208, | |
| "24.up_proj": 0.6403064727783203, | |
| "24.down_proj": 0.6378594636917114, | |
| "25.q_proj": 0.578160285949707, | |
| "25.k_proj": 0.5866066217422485, | |
| "25.v_proj": 0.6443796753883362, | |
| "25.o_proj": 0.6698061227798462, | |
| "25.gate_proj": 0.6242399215698242, | |
| "25.up_proj": 0.6427105069160461, | |
| "25.down_proj": 0.6289896965026855, | |
| "26.q_proj": 0.5467449426651001, | |
| "26.k_proj": 0.5561713576316833, | |
| "26.v_proj": 0.6325230002403259, | |
| "26.o_proj": 0.6914882659912109, | |
| "26.gate_proj": 0.6260568499565125, | |
| "26.up_proj": 0.6445198059082031, | |
| "26.down_proj": 0.6264070868492126, | |
| "27.q_proj": 0.5962280631065369, | |
| "27.k_proj": 0.6041812300682068, | |
| "27.v_proj": 0.6541686058044434, | |
| "27.o_proj": 0.6721190810203552, | |
| "27.gate_proj": 0.6289679408073425, | |
| "27.up_proj": 0.6461732983589172, | |
| "27.down_proj": 0.622887909412384, | |
| "28.q_proj": 0.577075719833374, | |
| "28.k_proj": 0.5881885290145874, | |
| "28.v_proj": 0.649710476398468, | |
| "28.o_proj": 0.6659877896308899, | |
| "28.gate_proj": 0.6287640333175659, | |
| "28.up_proj": 0.6446463465690613, | |
| "28.down_proj": 0.6314065456390381, | |
| "29.q_proj": 0.5415160059928894, | |
| "29.k_proj": 0.5502261519432068, | |
| "29.v_proj": 0.6414359211921692, | |
| "29.o_proj": 0.6688233613967896, | |
| "29.gate_proj": 0.6273680329322815, | |
| "29.up_proj": 0.6402621865272522, | |
| "29.down_proj": 0.6377818584442139, | |
| "30.q_proj": 0.5716490745544434, | |
| "30.k_proj": 0.5806272625923157, | |
| "30.v_proj": 0.650600016117096, | |
| "30.o_proj": 0.6727860569953918, | |
| "30.gate_proj": 0.6200348138809204, | |
| "30.up_proj": 0.6269094944000244, | |
| "30.down_proj": 0.635309100151062, | |
| "31.q_proj": 0.4421253800392151, | |
| "31.k_proj": 0.4497847557067871, | |
| "31.v_proj": 0.5590528845787048, | |
| "31.o_proj": 0.5637112855911255, | |
| "31.gate_proj": 0.5982303023338318, | |
| "31.up_proj": 0.600795328617096, | |
| "31.down_proj": 0.5228932499885559 | |
| }, | |
| "step": "swift_svd_dynamic_finetuned", | |
| "compression": { | |
| "upstream": "hiahei/Swift-SVD", | |
| "step": "swift_svd_dynamic", | |
| "alloc": "dynamic", | |
| "alpha": 0.6, | |
| "nsamples": 256, | |
| "seqlen": 2048, | |
| "dataset": "WikiText2", | |
| "seed": 42, | |
| "gqa_budget": null, | |
| "param_fraction": 0.4998633824481865, | |
| "mean_rel_err": 0.5765954015909561, | |
| "max_rel_err": 0.691561222076416 | |
| }, | |
| "factorisation": { | |
| "max_rel_err": 0.0014744423096999526, | |
| "factor_params": 3237117952, | |
| "dense_params": 6476005376 | |
| }, | |
| "finetune": { | |
| "protocol": "svdllm_stage2_u_then_v", | |
| "script": "gitclones/SVD-LLM/utils/LoRA.py", | |
| "data": "alpaca-cleaned", | |
| "epochs_per_half": 2, | |
| "learning_rate": 0.0001, | |
| "batch_size": 64, | |
| "micro_batch_size": 4, | |
| "lora_r": 8, | |
| "lora_alpha": 16, | |
| "lora_dropout": 0.05, | |
| "cutoff_len": 256, | |
| "val_set_size": 2000, | |
| "max_train_samples": "all", | |
| "precision": "bf16 (upstream hard-codes fp16; see run_with_torch_compat)" | |
| } | |
| } |